concurring in the result:
The court today does violence to two fundamental — and conservative — principles of our Constitution: Separation of Powers and Federalism. It does the first to no purpose whatsoever; it does the second needlessly, by reaching out to express views on an issue that is not yet before us. It is led to these results by an understandable desire to follow what it believes to be the will of Congress. But in so doing, the court attributes to the legislature views that Congress is most unlikely to have even considered, let alone held.1
This case raises principally two ques-. tions. The first is whether Congress, by the language it used in the PLRA, sought directly to terminate preexisting court-ordered consent decrees,2 or whether, instead, Congress used its paramount powers merely to alter the underlying law and left it up to the courts to exercise their traditional role of applying those changes. This question may seem formalistic, since the courts, responding to the' congressional action, would in the case before us be inexorably led either to terminate the decrees or to bar future relief under them, thereby reaching precisely the same result whichever reading is given to the congressional language. In fact, however, it entails issues of the most fundamental sort concerning the Separation of Powers.
*172The second question asks to what extent Congress intended by the PLRA to limit the ability of prisoners to enter into contracts, enforceable in state courts, by which the prisoners obtain benefits in exchange for the waiver of possible federal claims. The answer to this latter question is entirely independent of the answer given to the first question. For, regardless of whether the PLRA directly terminated (or indeed vacated) federal court judgments or merely led courts not to give these consent decrees future effect, it would still remain uncertain whether the PLRA permitted state courts to find in existence and to enforce, under state contract law, agreements arrived at by the parties contemporaneously with the federal court consent decrees. I respectfully believe that the court today answers both of these questions incorrectly. Because the answers to these questions do not affect the outcome in the case before us, and because, as to some of the subsidiary issues in the case, I agree with the majority’s reasoning as well as its conclusion, I concur in the court’s result,3 while strenuously, though respectfully, dissenting from most of its reasoning.
7. DOES THE PLRA DIRECTLY TERMINATE THE CONSENT DECREES?
A. The Language of the PLRA and Legislative Intent
The relevant provision of the PLRA, 18 U.S.C. § 3626(b)(2), reads:
Immediate Termination of Prospective Relief. — In any civil action with respect to prison conditions, a defendant or in-tervener shall be entitled to the immediate termination of any prospective relief if the relief was approved or granted in the absence of a finding by the court that the relief is narrowly drawn, extends no further than necessary to correct the violation of the Federal right, and is the least intrusive means necessary to correct the violation of the Federal right.
18 U.S.C.A. § 3626(b)(2) (West Supp.1998).
On its face, this language does not terminate the consent decrees or any other court judgments. It does no more than prohibit -prospective relief under the prior judgments. And, as both the panel opinion and the in banc majority concluded, such termination of prospective relief in federal courts is clearly constitutional. The in banc majority also holds that this language bars state courts from granting prospective relief under the consent decrees and that it does so constitutionally.
On this latter point, the panel had held the opposite. Despite some legislative history to the contrary, the panel concluded that the PLRA terminated federal court jurisdiction but left state courts free to enforce the decrees.4 The question is a close one, and I am not fully convinced by the majority, but it is sufficiently in doubt that I am not inclined to dissent on the issue. Moreover, I agree completely that, if the PLRA limits state as well as federal courts from granting prospective relief under the decrees, it does so constitutionally.
Were this all there was to this section of the PLRA, the in banc court and I would not diverge on the first of the two issues before us. Our disagreement stems from the way in which the majority of the court interprets the definitional section that applies to this part of the PLRA. That section reads as follows:
(1) the term “consent decree” means any relief entered by the court that is based in whole or in part upon the consent or acquiescence of the parties but does not include private settlements;
(5) the term “prison” means any Federal, State, or local facility that incarcerates or detains juveniles or adults ae-*173cused of, convicted of, sentenced for, or adjudicated delinquent for, violations of criminal law;
(6) the term “private settlement agreement” means an agreement entered into among the parties that is not subject to judicial enforcement other than the reinstatement of the civil proceeding that the agreement settled;
(7) the term “prospective relief’ means all relief other than compensatory monetary damages;
(9) the term “relief’ means all relief in any form that may be granted or approved by the court, and includes consent decrees but does not include private settlement agreements.
18 U.S.C. § 3626(g).
The majority takes this definitional section to mean that the federal court judgments — the consent decrees themselves— are part of the relief that the PLRA directly terminates. It concedes that such a reading, albeit literal, is not natural. See ante at 181. And, after correctly suggesting that Congress has a right to define one thing to mean something different from what it normally means, the court looks to the legislative history of the PLRA to conclude that Congress has, in effect, done just that. See ante at 184-85. In this respect, the court agrees with the First Circuit’s decision in Inmates of Suffolk County Jail v. Rouse, 129 F.3d 649 (1st Cir.1997), cert. denied, — U.S. -, 118 S.Ct. 2366, 141 L.Ed.2d 735 (1998), which, after finding the same reading of § 3626(g) to be unlikely, nevertheless, on the basis of legislative history, deemed it to be what Congress intended. See id. at 654.5
The problem with such a reading — apart from its admitted linguistic awkwardness — is that there is no reason to suppose that Congress intended it. The legislative history cited by the majority (and by the First Circuit) surely shows that Congress meant, and vehemently so, not to have such consent decrees enforced by federal courts except as strictly necessary to protect federal constitutional rights. See H.R. Conf. Rep. No. 104-378, at 166 (1995); H.R.Rep. No. 104-21, at 24-26 (1995). Less clearly, the same legislative history can be read — as the in banc majority does — to indicate that Congress, having *174barred prisoners from obtaining federal relief under the decrees, wished to prevent them from going into state court and getting the same decrees enforced there. See, e.g., H.R.Rep. No. 104-21, at 25 (noting that the provision will “insuref ] that inmates will not simply run from the federal courthouse to the state courthouse to bring the same suits and to demand the same burdensome and unnecessary relief that the federal courts have irresponsibly imposed on local judicial systems”). But there is nothing in what the majority (and the First Circuit) cites, or in what I have been able to find, that indicates that Congress meant to infringe on an existing judgment of the courts, rather than merely to change the underlying law, so that the courts would reach the same ultimate result by modifying the effect of their own prior rulings. In other words, there is nothing in the legislative history that in any way suggests that Congress cared a jot or tittle about whether relief under the decrees was to be precluded as a result of court adherence to the PLRA, or whether the prior court judgments — that is, the decrees themselves — were to be terminated directly by congressional fiat.6
The difference between the two, though seemingly of no interest to Congress— which wanted certain results and was understandably unconcerned with process— is fundamental to a sound doctrine of Separation of Powers and hence to the constitutionality of the PLRA. As the majority reads the PLRA, Congress would for the first time in our constitutional history have validly commanded the alteration or termination of a court judgment. Giving the statutory language its more natural meaning would instead read the law to do no more than what Congress (with the courts’ approval) has frequently done in the past. See, e.g., Pennsylvania v. Wheeling & Belmont Bridge Co., 59 U.S. (18 How.) 421, 431-32, 15 L.Ed. 435 (1855) (“Wheeling Bridge II ”). And this still would achieve precisely the same results in terms of the prisoners before us and of their rights under these decrees. I believe the court reaches its radical conclusion because it fails to appreciate the appropriately formalistic nature of the Separation of Powers doctrine.
B. Good Fences Make Good Neighbors 7
The Sovereign cannot, either as a result of contracts and consent decrees into which it has entered, or as a result of court judgments, be precluded from governing in the public interest to the full extent of its powers. Thus, a contract by which the government agrees not to go to war for a certain number of years in exchange for an agreement by a manufacturer to produce pruning hooks instead of spears cannot *175keep the Sovereign from initiating hostilities, and neither can a consent decree between the manufacturer and the Sovereign (entered into, let us say, in settlement of an antitrust action in the spear industry). Similarly, a consent decree (or a contract, for that matter) by which a state agrees to furnish cigarettes to prisoners if they are willing to drop an action against the state cannot keep the government (either state or national) from passing laws prohibiting the distribution of cigarettes. The fact that in all such cases no court would thereafter be able to enforce the preexisting contract or the prior court judgment (embodied in a consent decree) is of no significance and does not, under the Separation of Powers doctrine, limit the capacity of the state validly to enact such laws.8
In the same way, the existence of court judgments that have some future consequences — orders that damages be paid, or that a nuisance or a continuing trespass be enjoined, for example — in no way precludes the Sovereign from altering laws governing legal tender or nuisance or trespass so that the courts will be led to ignore or to alter their earlier rulings and thereby, in effect, to nullify their prior judgments. This is so with respect to judgments that have predominantly present effects. Thus, a court would necessarily and properly amend a decree ordering that damages be paid in gold so as not to violate a subsequent statute prohibiting the private transfer of gold. This is also so, and more likely to happen, when a judgment (like a consent decree or an injunction) has important future effects.
The difference between the two situations is simply that, in the latter case, the probability that the legislature will intervene — and render the prior judgment, contract, or consent decree effectively unenforceable by the courts — is much greater than in the former. In both cases, however, the point is the same. The Sovereign being Sovereign cannot be kept from passing laws, and thereby furthering its view of the public interest, merely because a contract, consent decree, or judgment of any sort exists and would be negatively affected or rendered unenforceable as a result of such legislation. The legislature, in all such eases, acts on its side of the constitutional fence. And its good neighbor — the judiciary — responds accordingly on its own terrain.
That is the meaning of System Federation No. 91 v. Wright, 364 U.S. 642, 81 S.Ct. 368, 5 L.Ed.2d 349 (1961), in which the Supreme Court required the “ ‘court of equity to modify an injunction in adaptation to changed conditions.’” Id. at 647, 81 S.Ct. 368 (quoting United States v. Swift & Co., 286 U.S. 106, 114, 52 S.Ct. 460, 76 L.Ed. 999 (1932)). In System Federation, the existing consent decree forbade party railroads and unions from discriminating against nonunion laborers. See id. at 644, 81 S.Ct. 368. A subsequent congressional amendment to the Railway Labor Act permitted, in certain circumstances, contracts requiring union labor. See id. As a result, the Supreme Court found that the district court had abused its discretion when it refused to modify the consent decree. See id. at 650-53, 81 S.Ct. 368; see also Swift & Co., 286 U.S. at 114, 52 S.Ct. 460 (“A continuing decree of injunction directed to events to come is subject always to adaptation as events may shape the need.”).
It is also the meaning of Wheeling Bridge II. In Pennsylvania v. Wheeling and Belmont Bridge Co., 54 U.S. (13 How.) 518, 14 L.Ed. 249 (1851) (“Wheeling Bridge I”), the Court had found that a bridge over the Ohio River was a public nuisance because its height obstructed the free navigation of the river and thus con*176travened regulations of Congress issued pursuant to the Commerce Power. See id., 54 U.S. (13 How.) at 626. A subsequent enactment by Congress had declared the bridge to be a post-road — and therefore not subject to nuisance law — and found that, at its proposed height, the bridge did not interfere with any public rights of free navigation. See Wheeling Bridge II, 59 U.S. (18 How.) at 426. Accordingly, in Wheeling Bridge II (which came back to the Court when the prior victors sought a contempt order against the bridge company), the Court recognized the change in the underlying law (and “rights”) and therefore determined that the injunctive decree could no longer be enforced by the courts. See id., 59 U.S. (18 How.) at 431-32.9
The language that the Court used in Wheeling Bridge II is instructive:
[T]hat part of the decree, directing the abatement of the obstruction, is executo-ry, a continuing decree, which requires not only the removal of the bridge, but enjoins the defendants against any reconstruction or continuance. Now, whether it is a future existing or continuing obstruction depends upon the question whether or not it interferes with the right of navigation. If, in the meantime, since the decree, this right has been modified by the competent authority, so that the bridge is no longer an unlawful obstruction, it is quite plain the decree of the court cannot be enforced. There is no longer any interference with the enjoyment of the public right inconsistent with law, no more than there would be where the plaintiff himself had consented to it, after the rendition of the decree.
Wheeling Bridge II, 59 U.S. (18 How.) at 431-32.
In these cases, and many others, see, e.g., Agostini v. Felton, 521 U.S. 203, -, 117 S.Ct. 1997, 2018, 138 L.Ed.2d 391 (1997) (ordering the district court to vacate “a continuing injunction entered some years ago in light of a bona fide, significant change in subsequent law”); Rufo v. Inmates of Suffolk County Jail, 502 U.S. 367, 388, 112 S.Ct. 748, 116 L.Ed.2d 867 (1992) (“A consent decree must of course be modified if, as it later turns out, one or more of the obligations placed upon the parties has become impermissible under federal law.”), the legislative and executive branches acted, within their powers, to alter the law, and the courts inevitably and properly adjusted their prior decrees to make the judgments conform to the law as it had come to be. In none of these cases, however, did the political branches themselves attempt to modify a court judgment. And in none of them did they order a court to do so. The fact that these branches undoubtedly wished to bring about the result achieved, and that they enacted the law with the expectation that the courts would inexorably be led to alter their prior rulings, in no way undermined the validity of what the political branches did ... on their side of the constitutional line.
*177When, instead, the legislature has sought on its own to modify a court decree, when it has ordered the alteration or reopening of a final judgment, the Supreme Court has expressly barred it from doing so. That is precisely what Plaut v. Spendthrift Farm, Inc., 514 U.S. 211, 115 S.Ct. 1447, 131 L.Ed.2d 328 (1995), told us is forbidden by the doctrine of Separation of Powers. As the Court stated in Plaut, “We know of no previous instance in which Congress has enacted retroactive legislation requiring an Article III court to set aside a final judgment, and for good reason. The Constitution’s separation of legislative and judicial powers denies it the authority to do so.” 514 U.S. at 240, 115 S.Ct. 1447; see also Hayburn’s Case, 2 U.S. (2 Dali.) 408, 411, 1 L.Ed. 436 (1792) (opinion of Wilson and Blair, JJ., and Peters, D.J.) (noting that Congress cannot interfere with the judicial determination that a claimant has a right in the future to receive a pension); id., 2 U.S. (2 Dali.) at 413 (opinion of Iredell, /., and Sitgreaves, D.J.) (same).10
The distinction is simple and directly germane to the case before us. If the PLRA constitutionally deprives the federal (and state) courts of jurisdiction to enforce consent decrees of the sort before us, then the courts affected will necessarily refuse to give future effect to these decrees. If instead the PLRA seeks directly to order the same courts to alter, modify, or terminate the decrees themselves, then it attempts to do exactly what has never been permitted. For then it grants the legislative and the executive branches naked power over the courts and their holdings. Such a grant impermissibly crosses the fence and trespasses on the judicial terrain.
In upholding this unprecedented infringement, the in banc court (like the other circuits that have read the PLRA directly to terminate the consent decrees) relies on dicta in Plaut. The Court in Plaut, upholding the strictest form of Separation of Powers, differentiated Wheeling Bridge II from the case then before it. It stated that Wheeling Bridge II distinguished itself, since it involved legislation that only “altered the prospective relief of injunctions,” rather than a current judgment. Plaut, 514 U.S. at 232, 115 S.Ct. 1447. I believe that the majority misreads this statement when it says that legislatures may directly terminate judgments so long as they have only future effects.
It is hard to see why it should make any difference — for the doctrine of Separation of Powers — whether the judgment of the judicial branch that is' legislatively interfered with is one that has prospec*178tive effect or one that does not. In both instances, the other branches directly invade judicial territory. Nor is such a distinction needed, or even useful, in explaining Wheeling Bridge II. The difference, as previously noted, between present and future effects can be crucial when the legislative and executive branches have not sought to alter a judgment but have instead promulgated a law that alters the legal rules on which the judgment was based. In such cases — for example, Wheeling Bridge II — it is clearly to be expected that a court will modify a judgment’s prospective effect. And, since there are no impediments to its doing so, that judicial modification is inevitable. See, e.g., System Federation, 364 U.S. at 647, 81 S.Ct. 368 (requiring such a change); Swift & Co., 286 U.S. at 114, 52 S.Ct. 460 (same). Conversely, it is rare that the courts have the authority to alter a judgment that has no future consequences. See, e.g., Agostini, at -, 117 S.Ct. at 2018 (“Intervening developments in the law by themselves rarely constitute the extraordinary circumstances required for relief under Rule 60(b)(6).... ”). But when they do, the rule of Wheeling Bridge II would apply to these judgments as well. See Pioneer Investment Services Co. v. Brunswick Associates Ltd. Partnership, 507 U.S. 380, 393-94, 113 S.Ct. 1489, 123 L.Ed.2d 74 (1993) (citing with approval Klapprott v. United States, 335 U.S. 601, 613-14, 69 S.Ct. 384, 93 L.Ed. 1099 (1949) (opinion of Black, J.) (permitting a judgment that had become final four years earlier to be reopened under Rule 60(b) in the interest of justice)); Matarese v. LeFevre, 801 F.2d 98, 106 (2d Cir.1986) (“A postjudgment change in the law having retroactive application may, in special circumstances, constitute an extraordinary circumstance warranting vacation of a judgment.” (emphasis added)).
All of these are totally different from the situation that occurred in Plaut (and not in Wheeling Bridge II). In Plaut, the legislative and executive branches purported to modify a judgment on their own. As such, they manifestly interfered with the judicial branch, and endangered the Separation of Powers. That, moreover, would have been so even if the judgment they abused implicated future instead of present effects. For then too, Congress and the President, by directly altering an order of the court, would have impermissibly invaded the judicial terrain. And that is precisely what Plaut says they may not do.11
*179The majority today seeks to blur the distinction by holding explicitly what most of the other circuits have held at least implicitly: that Congress may only directly affect an existing injunctive decree if the legislature both has the power to alter the underlying substantive law and has actually exercised that power in the statute at issue. See ante at 188. This express limitation is highly desirable in that it precludes Congress from seeking to modify prior judgments both in areas as to which it has no underlying constitutional authority (for example, diversity cases), and also in areas where it has the authority, but has not consciously chosen to exercise it. The majority's limitation, however, in no way cures the fundamental Separation of Powers defect. It remains the fact that, under the majority’s ruling, the legislature is, for the first time in our history, permitted directly to order courts to modify their final judgments. And the court fails to explain why such a direct infringement is more acceptable when dealing with judgments that have long-term future effects (like injunctions) than it is as to judgments the effects of which are more proximate (like restitution or damages).
The majority’s approach, moreover, highlights the manifest absurdity of ascribing to Congress the intent to affect directly the decrees in this case. The majority’s holding means that in order to command direct termination of the decrees, Congress must have changed the underlying law so as to render the decrees unenforceable. But once it has done so — and we all agree that it has — the Supreme Court’s jurisprudence, as set forth in cases like System Federation, 364 U.S. 642, 81 S.Ct. 368, 5 L.Ed.2d 349 (1961), and Swift & Co., 286 U.S. 106, 52 S.Ct. 460, 76 L.Ed. 999 (1932), makes clear that lower courts are themselves obligated to comply with the altered underlying law and to cease to enforce the decrees. Why on earth then would Congress intend to do more and seek to void the decrees directly when the same result would occur anyway? . And why should we read the legislature’s admittedly awkward language to do this when such a reading has radical consequences for the traditional division between courts and legislatures? Indeed, the fact that the majority relies upon System Federation and Swift & Co. to make its point, see ante at 188-89, drives home the fact that holding that Congress terminated the decrees themselves (rather than that it altered the underlying law) does not affect the practical result in this case. It merely allows Congress gratuitously to undermine the doctrine of the Separation of Powers and invade what is traditionally the province of the judiciary.
C. The Formal Structure of Separation of Poivers Doctrine
The distinction between altering the law underlying a judgment — thus leading the courts to modify a judgment — and altering the judgment itself may seem a formalistic one. But that is not a valid criticism.12 As Justice Scalia, the author of the majority opinion in Plaut, wrote in another context, “Of all the criticisms leveled against textualism, the most mindless is that it is ‘formalistic.’ The answer to that is, of course it’s formalistic! The rule of law is *180about form.... Long live formalism. It is what makes a government a government of laws and not of men.” Antonin Scalia, Common-Law Courts in a Civil Law System: The Role of United States Federal Courts in Interpreting the Constitution and Laws, in A Matter of Interpretation: Federal Courts and the Law, at 3, 25 (Amy Gutmann ed., 1997) (emphasis in original). His statement applies equally here.13
The Separation of Powers doctrine, Plant tells us, is a “structural safeguard rather than a remedy to be applied only when a specific harm, or risk of specific harm, can be identified.” Plaut, 514 U.S. at 239, 115 S.Ct. 1447. And the doctrine is a “prophylactic device, establishing high walls and clear distinctions because low walls and vague distinctions will not be judicially defensible in the heat of inter-branch conflict.” Id.
It is no accident, therefore, that what some may label as “formalisms” are a mainstay of the Court’s Separation of Powers and Federalism decisions. The Court’s recent decision in Printz v. United States, 521 U.S. 98, 117 S.Ct. 2365, 138 L.Ed.2d 914 (1997), reiterated this point:
Much of the Constitution is concerned with setting forth the form of our government, and the courts have traditionally invalidated measures deviating from that form. The result may appear “formalistic” in a given case to partisans of the measure at issue.... But the Constitution protects us from our own best intentions: It divides power among sovereigns and among branches of government precisely so that we may resist ... temptation....
Id. at -, 117 S.Ct. at 2383 (quoting New York v. United States, 505 U.S. 144, 187, 112 S.Ct. 2408, 120 L.Ed.2d 120 (1992)) (internal quotation marks omitted).
In other words, when great structural boundaries are at stake, the Court has repeatedly made clear that hoiv something is done (i.e., whether a judgment is directly altered or whether the underlying law is changed in such a way as to lead the courts to modify their judgments) can be at least as important as the result that is achieved.14
In the context of the Separation of Powers, it is the formal independence of the courts that is safeguarded by an absolute rule — never before infringed — that says that court judgments are not to be touched by the legislative and executive branches. The basic power to legislate (and to enforce that legislation) in the public interest is not undercut by this formal rule (that is, the power to achieve a result in terms of prisoners’ rights is unaltered), but the dignity of the courts qua courts is protected by it.15
*181 D. What the PLRA Intended To Do
Thus, if we were to conclude that Congress intended by the PLRA directly to terminate the consent decrees, and thereby on its own to modify judgments, we' would have to face the question of whether such a radical step was constitutionally permitted despite its patent infringement of the line that divides the courts from the other branches. But, as we have seen (as the in banc majority itself admits, and as the panel argued at length), the language of the statute requires no such interpretation. See ante at 183-85; Benjamin, 124 F.3d at 167. Is that reading nonetheless mandated by the legislative history of the PLRA? I think not.
Let me be clear that I am not, in this respect, simply relying, as did the panel, on the mass of cases that suggest (properly) that statutory language should be read to avoid serious constitutional difficulties,16 though those cases would certainly be enough to justify my reading. In the case before us, the argument against the existence of a congressional intent to terminate the consent decrees finds a far simpler grounding. If — as I believe can be demonstrated — everything that Congress sought to achieve by this part of the PLRA is as well accomplished by terminating the future effects of the decrees as by destroying the judgments themselves, it is ill-advised to suppose that Congress, by its ambiguous language, meant to do something that raises such severe Separation of Powers problems.
Before one assumes that, regardless of results, Congress intended to be so radical, one must, I suggest, have some direct indication that the legislature wanted specifically to infringe on the dignity of the courts, and to do so despite the formal requirements of the Separation of Powers doctrine. One must, in other words, have a clear sign that it did not merely wish to *182achieve certain ends with respect to prisoner litigation. Yet, neither the in banc court, the First Circuit, nor any other court that has dealt with these provisions of the PLRA has cited anything that Congress might have tried to accomplish by this part of the PLRA that is not obtained precisely to the same degree regardless of whether the decrees are themselves terminated or, instead, the federal and state courts are denied jurisdiction to give the decrees future effect.
The in banc majority argues at length that Congress meant to get the federal courts out of the business of running jails, and it cites any number of congressional statements to that effect. I agree. But whether the consent decrees before us are terminated (or, for that matter, vacated) makes no difference whatsoever in this regard, so long as it is the case that federal courts are denied jurisdiction to give future effect to the decrees. And that is precisely what will occur under my alternate reading of the statute.17
The majority states that Congress sought to preclude state courts — as well as federal ones — from giving future effect to the decrees. It therefore interprets the PLRA as preventing a prisoner (barred from getting enforcement of the consent agreement in federal court) from going to state court and getting the decree enforced there. See ante at 182-83. But, once again, the posited congressional aim is as readily accomplished by giving the statutory language its natural meaning so that the PLRA (by limiting the jurisdiction of state as well as federal courts) terminates future relief under the decrees as it is by ending the decrees themselves.
There remains the question of whether Congress intended to prohibit state contract suits designed to vindicate prisoner claims that might also be recognized in the decrees. And that question, which is the second issue as to which I disagree with the majority, is a complicated one about which I will have more to say anon. For purposes of the issue that I am currently addressing, however, the question can be dealt with quickly. The survival of state law contractual rights vel non is unaffected by whether the decrees themselves are terminated or whether, instead, future relief under the decrees is forbidden in both state and federal courts. The issue of whether private state contractual rights perdue remains the same on the reading that would end the decrees themselves and on the reading that would not directly touch the judgments but would bar federal and state courts from granting future relief under them.18
It follows that it makes no sense to assume that Congress intended under the PLRA to do something never before done. Since everything it wished to achieve is as fully accomplished by less radical methods, courts should not just hesitate, they should be bound to refrain, from giving the congressional language an awkward (though not impossible) reading, which, having no purpose, constitutes a gratuitous contravention of the boundaries established by the Separation of Powers doctrine.
E. Are the Limitations Imposed by the PLRA on State and Federal Court Jurisdiction Constitutional?
Before I turn to the second issue in this case — the survival of state law contractual rights under the PLRA — I must briefly address the question of whether Congress can constitutionally do what I read the PLRA as doing, namely limit severely the jurisdiction of both federal and state courts so as to bar most prospective relief under the consent decrees involved in the *183case before us. If Congress can so restrict the courts’ jurisdiction, then — as I have just argued — the same results are achieved under my reading of the statute as under that of the in banc majority. If not — if the only way Congress can control the actions of state or federal courts is by terminating the decrees themselves — then my argument that Congress could not have intended the radical step of directly altering a court judgment would be seriously undercut. Fortunately, there can be no doubt that the restrictions on both federal and state court jurisdictions that I believe the PLRA mandates are constitutionally valid.
Whatever may be the outer boundaries of Congress’s power to restrict the jurisdiction of the lower federal courts,19 there is no question that the power of such courts can be limited to cases involving federal rights. See Lockerty v. Phillips, 319 U.S. 182, 187, 63 S.Ct. 1019, 87 L.Ed. 1339 (1943); Lauf v. E.G. Shinner & Co., 303 U.S. 323, 330, 58 S.Ct. 578, 82 L.Ed. 872 (1938). The PLRA affects only cases in which there is no finding that the “relief is ... necessary to correct the current and ongoing violation of the Federal right” by “the least intrusive means” possible. 18 U.S.C. § 3626(b)(3). Hence the validity of such a restriction of jurisdiction cannot be seriously questioned. See ante at 189-91; Benjamin, 124 F.3d at 169-70.
The issue is only slightly more complicated with respect to state courts. What, one might ask, is the constitutional basis for a federal limitation on the jurisdiction of state courts? The answer is that the PLRA prohibits state courts from giving effect to relief only when that relief is embodied in consent, decrees that were entered into in settlement of suits claiming federal rights. Because the suits that culminated in the federal court decrees involved such “national” rights, Congress both directly and under the Necessary and Proper Clause can bar state courts from giving future relief under them. The original claims that gave rise to the decrees were federal ones, and Congress, which has authority over such claims, surely can restrict any court from giving future relief in exchange for their relinquishment. The question is, therefore, not one of federal power — the power is there. It is one of whether Congress intended to use that power. And the majority holds that Congress did indeed intend such a limitation.20
I conclude, therefore, that everything that Congress, by its language and by its intent as expressed in the legislative history of the PLRA, wished to accomplish can be constitutionally effectuated without reading the statute to order a direct infringement of court judgments.21 Since *184the statute in no sense requires such a reading, and since such a reading can properly be viewed as “awkward” and “not natural,” it seems to me extraordinarily misguided so to interpret the Act. In doing this, the in banc majority (and the other circuits that have done the same) permits the constitutional walls between the judiciary and the other branches of government to be breached — and does so for no reason. A panel of the Ninth Circuit gave the same reading in order to strike the statute down.22 It is hard to know which approach is more damaging to the genuine will of the Congress and to our constitutional structure.
II. DO PRIVATE STATE CONTRACT CAUSES OF ACTION SURVIVE THE PLRA1?
The second issue that divides me from the in banc majority is whether a state court can (after passage of the PLRA) determine that the prisoners before us, at the same time that they entered into consent decrees, also reached a meeting of minds with the entities that they had sued and that this meeting of the minds constituted a contract that, in the normal course of things, could be enforced in state courts. The question is properly divided into three parts: (1) Can parties to a consent decree be said under state law to have, independently of that decree, entered into a valid contract — at least in some circumstances? (2) If such a contract would — in some instances — be valid under state law, does the PLRA prohibit its enforcement in state courts? (3) What, if any, restrictions would a finding that valid state contracts existed place on future actions by state legislators with respect to such contracts?
I contend that the in banc majority’s answer to the first question is probably wrong and, more important, improperly attempts to decide a question that is purely one of state law, thereby offending the most fundamental principles of Federalism. The answer to the second question is more complicated. Congress surely has the constitutional authority to prohibit the enforcement of otherwise valid state contracts when such contracts are entered into in circumstances like those before us. But, on the language and history of the PLRA, it is unlikely that Congress intended to prohibit such state law contracts. *185More fundamentally, since the issue is not before us in the current case, we cannot and should not decide it.23 The answer to the third question is that very few restrictions would impede state legislatures from modifying such contracts.
A. Can Parties to a Consent Decree Enter into a Separate and Potentially Valid State Lato Private Contract?
The question of whether the parties, at the same time that they concluded the consent decrees, also achieved a meeting of minds such that a state might deem a state law contract to have been established, is — by definition — a state law issue. It is a totally different question from whether a federal consent decree can itself be enforced in state courts. By conflating the two issues, the court permits itself to cast doubt on whether such a requisite meeting of the minds can exist. But that is not for a federal court to say. What New York deems a valid contract is for New York to decide. And for federal courts to instruct New York on what does or does not constitute a proper meeting of the minds under New York law is as grievous a crossing of the boundaries imposed by Federalism24 as the direct legislative alteration of a court judgment is of the boundaries imposed by the Separation of Powers.
Moreover, it would not be surprising if New York determined that at least some of the many consent decrees whose future effects the PLRA seeks to terminate did in fact give rise to state law contractual rights. In this respect, it is important to realize that the consent decrees that may be affected by the PLRA are many and highly varied in New York alone. A few— like the ones in the- case before us — have generated a huge cavalcade of judicial administration and contain terms that were the result of continuous evolution and articulation over many years.25 Many others simply settled a claim by giving the prisoner claimants a particular future benefit, and have lain judicially dormant ever since. To make a one-size-fits-all abstract judgment with respect to such differing agreements, as the majority appears to do, seems wrong to me. And it would be wrong even if the majority had the author*186ity to do it.26
In the end, though, the essential point remains the same.' For federal courts to tell states what they may, or should, or might consider to be contractually enforceable rights is simply not proper. The in banc court’s opinion seems to do this, and I can only conclude that in its understandable eagerness to make clear that the consent decrees themselves are not enforceable in state courts, it went beyond what it can appropriately decide without perhaps intending to do so.27 Moreover, since we cannot rule on what is outside our authority to determine, I believe that the statements by the in banc court with respect to the existence of such state law contracts, necessarily, have no legal effect.
B. Does the PLRA Constitutionally Ban the State Courts’ Enforcement of Contracts That Would Otherwise Be Valid Under State Law?
Holding that state law must decide whether a contract (enforceable in state courts) could have been created by the same circumstances that gave rise to the consent decrees, however, does not resolve the issue before us. We must also ask— even if the conduct of the parties gave rise to a state law contract — whether that agreement was made unenforceable by the PLRA. And this question itself divides into three parts: (1) whether federal law can constitutionally bar enforcement of otherwise valid state law contracts like the ones before us, (2) whether it did so through the PLRA, and (3) whether this question is properly before us.
The answer to the first question is an easy yes. For the same reasons that federal law can bar state courts from giving future effect to these consent decrees, it can also bar enforcement in state courts of contracts that have arisen, if at all, in circumstances like those involved in the instant case. The contracts that state law might deem to have been made are all agreements for which the alleged consideration was the giving up of federal claims (and usually federal constitutional claims). There can be no doubt that Congress — if, for instance, it wished to discourage parties from too readily abandoning federal constitutional claims — could validly make unenforceable all contracts that waived or otherwise gave up such claims. One may question if that is what the Congress that passed the PLRA had in mind — but that is another issue. What one cannot deny'is that the federal government can control *187the abandonment of federal claims and can do so, if it wishes, by barring even state court enforcement of contracts that were negotiated in exchange for those claims.
The issue then becomes whether Congress did so under the PLRA. On the whole, I do not believe that it did. But, ultimately, this is not a question that is properly before us in this case, and we should therefore not attempt to answer it definitively now.
Whether Congress did so or not depends on the reading that is given to two seemingly contradictory clauses of the PLRA. The PLRA states:
(c) Settlements.—
(1) Consent decrees. — In any civil action with respect to prison conditions, the court shall not enter or approve a consent decree unless it complies with the limitations on relief set forth in subsection (a).
(2) Private settlement agreements.—
(A) Nothing in this section shall preclude parties from entering into a private settlement agreement that does not comply with the limitations on relief set forth in subsection (a), if the terms of that agreement are not subject to court enforcement other than the reinstatement of the civil proceeding that the agreement settled.
(B) Nothing in this section shall preclude any party claiming that a private settlement agreement has been breached from seeking in State court any remedy available under State law.
18 U.S.C. § 8626(c).
Paragraph (c)(2)(A) appears to permit a private settlement agreement only “if the terms of that agreement are not subject to court enforcement other than the reinstatement of the civil proceeding that the agreement settled.” As such it would seem to bar state law contract remedies, since these typically go well beyond permitting the reinstatement of the suit settled by the contract.
Paragraph (c)(2)(B), however, immediately limits (c)(2)(A) and states “[njothing in this section- shall preclude any party claiming that a private settlement agreement has been breached from seeking in State court any remedy available under state law” (emphasis added). And this paragraph clearly seems to indicate that all state law remedies remain available, including those that go beyond a reinstatement of the suit that the contract sought to settle. The. very next paragraph of the statute, moreover, states: “[t]he limitations on remedies in this section shall not apply to relief entered by a State court based solely upon claims arising under State law.” Id. § 3626(d).
How is one to reconcile these provisions? It seems likely that paragraph (c)(2)(A), when it limits court enforcement to “the reinstatement of the civil proceeding that the agreement settled,” id. § 3626(c)(2)(A), is referring to federal court enforcement. In the prior paragraph the “court” whose power to grant a consent decree is being restricted manifestly refers to the federal court. And when succeeding paragraphs speak of the right to receive any remedy available under State law, the “court” in which these remedies can be obtained is instead expressly described as a “State court.” See, e.g., id. § 3626(c)(2)(B).
For these reasons, I believe the better reading of these provisions is likely to be that the PLRA, whatever it meant to do to the consent decrees themselves or to future relief under them, did not also intend to prohibit or limit state contracts made in settlement of federal claims. It follows, therefore, that if a state court were to deem the meeting of minds that led to the consent decrees to be sufficient to give rise to valid state law contracts, the PLRA does not appear to bar local law enforcement of such local law claims. It does not seem to bar it, not because Congress could not do so constitutionally, but because Congress appears not to have precluded the suits.
*188In the end, though, this is a question that we should not decide at this time. We do not know whether any, some, most, or none of the consent decrees (whose future effects the PLRA sought to limit) arose in situations out of which state courts will find valid state law contracts also to have been made. Unless and until a state court finds that a valid state law contract exists, it is premature for us to rule on whether the PLRA would bar the local courts from enforcing such a contract. Since, moreover, the parties seeking relief before us have not asked for any preclusion of state court contract claims — they only seek termination and vacatur of the decrees — it would be improper for us to rule on the issue even if it were not premature. Once again, the question is not before us.28
There remains one important point to be made with respect to state law contract claims. It is that the existence and survival of such state law claims under the PLRA is completely independent of whether the consent decrees are vacated, terminated, or left in place (with both federal and state courts barred from giving relief under them). If (as several circuits have held) the PLRA vacates the decrees, this' in no way means that state courts cannot find that a valid contract (independent of the decrees) had been formed and was enforceable under state law. Indeed, such a state law finding would seem to fit precisely under the saving clause of § 3626(d) since by definition once the decree was vacated the claim could only be “based solely upon ... State law,” given that no federal claims would be left. And the same would be true if the decrees were terminated, as ordered by the in banc (and First Circuit) holding. Once again any “relief entered by a state court” would have to be “based solely upon [contract] claims arising under State law.” For here too, once the decrees were terminated, the claims would have no other basis.
It might be tempting to say that under the interpretation of the PLRA that bars future relief under the decrees but does not terminate them, the validity of state contract claims is, if anything, more doubtful. After all, if the decrees remain extant, it might seem at least plausible that the “relief entered by a State court” was not “based solely upon claims arising under State law” (and hence be permitted) but could, in part, derive from the still unterminated decrees (and hence be barred). But, in fact, that is not possible. On this reading, the PLRA bars all relief under the decrees (in state as well as federal courts). It follows that, also on this interpretation, any “relief entered by a State court” must be “based solely upon claims arising under State law.”
Thus, nothing in the discussion of the possible existence of state law contracts alters what I said earlier when considering the Separation of Powers issue. Since the existence and survival of state law claims is independent of whether the decrees are obliterated or are left in place (and only prospective relief is barred), precisely the same result obtains whichever reading is given to the PLRA.
*189 C. What if Any Restrictions Would a Finding of Valid State Contract Claims Impose on State Legislatures?
At oral argument, it was suggested that, if we were to hold that state law contracts could continue to give the relief that was no longer available under the consent decrees, states would be forever frozen into agreements into which they had foolishly entered, even when those agreements were no longer in the public interest. Such an argument, however, misperceives the degree to which a contract binds the state as Sovereign.
As I said at the beginning of this opinion, the Sovereign cannot contract away its powers to govern. It cannot do that any more than it can be precluded from governing by a court judgment or by a consent decree. If the public interest requires it, the Sovereign can pass laws in derogation of any of these, subject only to the constitutional limits that apply to its lawmaking generally. Thus, the federal government can act in derogation of contracts that it has entered into, so long as it is operating under one of its paramount powers and so long as it does not violate any constitutional prohibition, such as those of the Bill of Rights. Similarly, the states can pass laws limiting the effect of their own contracts so long as, in doing so, they do not violate any express constitutional prohibitions or valid federal laws.
The fact that the Sovereign is itself a party to a contract is not irrelevant, but it is also in no way determinative. Courts will look with greater skepticism when the Sovereign is a party to a contract and seek to determine whether the government’s actions are those of a contracting party trying to get out of a financially bad deal or whether, instead, the government is acting as Sovereign in the public interest. See Calabresi, supra, 71 Yale L.J. at 1203.
This, in the case of the federal government, means that the courts will not seek out a paramount power that would validate the government’s action, when that action enables the government to renege on an unfavorable contract. But it also means that the federal government can nonetheless abrogate a contract that is financially disadvantageous to it, if the statute doing so points out the paramount power and the public interest served by the abrogation. See Resolution Trust Corp. v. Federal Sav. & Loan Ins. Corp., 25 F.3d 1493, 1501 (10th Cir.1994) (noting that where the federal government is a party to a contract, a statute modifying the contract will be subjected to the “sovereign acts” doctrine, under which Congress may abrogate contracts through general legislation, but may not target a class of contracts to which it is a party); see also United States v. Winstar Corp., 518 U.S. 839, 898, 116 S.Ct. 2432, 135 L.Ed.2d 964 (1996) (Souter, ./.) (plurality opinion) (“The greater the Government’s self-interest, ... the more suspect becomes the claim that its private contracting partners ought to bear the financial burden of the Government’s own improvidence.... ”); National R.R. Passenger Corp. v. Atchison, Topeka & Santa Fe Ry. Co., 470 U.S. 451, 472, 105 S.Ct. 1441, 84 L.Ed.2d 432 (1985).
The same reasoning applies, mutatis mutandis, to the states, whose sovereignty is both broader and more constricted than that of the federal government. The not unduly burdensome restrictions of due process remain and restrain both Sovereigns to some degree. But, in the end, it should be clear that even if state contract law recognizes a valid agreement between the parties before us, and the PLRA does not itself bar enforcement of the agreement in state courts, this in no way precludes the state of New York, acting as Sovereign in the public interest — and not solely as a contractor looking after its pecuniary interest — from modifying or even doing away with the contract.
III. A CAUTIONARY NOTE
Having concluded first that the PLRA does not terminate the consent decrees *190themselves but only ends future relief (either in state or federal courts) under them, and second that the state law of contracts might give relief analogous to that which is no longer available under the decrees and that such relief is not obviously barred by the PLRA, I am bound both by comity and the respect that I feel for my siblings to give weight to the fact that my views have found no adherence among the other circuits and have been forcefully rejected in my own court. I have done so. But in the end, I still believe that extremely able judges with the best of intentions have decided these issues incorrectly and that there are plausible reasons explaining why this has happened. It is to those reasons that I turn in concluding this opinion.
The first reason is that we all have been trained to downplay the significance of formal rules and to concern ourselves predominantly or even exclusively with results. We tend to look at law primarily as an engine to achieve immediate goals, and to ignore its symbolic significance. Thinking this way, it is all too easy to equate (a) the termination of judicial judgments with (b) changes in underlying laws, when both bring about the same results. But in fact we are wrong.
Of course, results do matter too. But they are not the only things that count. Symbols and formal boundaries can also be deeply important. And an overly functional view of the law can lead to grievous errors. For example, are distinctions expressly based on race the same as rules that have disparate impact on different racial groups? Of course not. This is not to say that disparate impact may not, in some circumstances, be constitutionally or legislatively prohibited.29 And, indeed, I have urged a greater recognition of the potential constitutional frailty of some arrangements that create disparate impacts. See Guido Calabresi, Foreword: Antidis-crimination and Constitutional Accountability (What the Bork-Brennan Debate Ignores), 105 Harv. L.Rev. 80, 115 n.106 (1991).30 But the fact that some disparate results may be invalid, regardless of how they are achieved, does not mean that the converse is true. It in no way suggests that, even when results can withstand constitutional scrutiny, there are not some ways of achieving the same goals that are barred by the great Charter. The means of accomplishing the ends — whether they be expressly race-conscious distinctions, suits for damages directly brought against the states, or legislative abrogations of court judgments — may fall in and of themselves, because to permit them is to allow something whose symbolism is intolerable, and remains so even when an identical result, achieved in other ways, may perhaps be permitted.
Nowhere is this more true than with respect to Separation of Powers. Despite the Supreme Court’s ringing formalistic language in this regard, functionalist judges — as all of us tend to be — are reluctant to follow the Court’s lead. And this *191carries particularly harmful consequences when Separation issues are involved. For, in this area, we cannot rely on any other institution of government to guard the formal boundaries. It is illusory to expect that the legislature or the executive will mind the fence, especially when it is the judiciary’s symbolic terrain that is trespassed upon. It is equally illusory to count on the parties to care about formal Separation of Powers notions. When these notions do not affect the result, the parties will not be much interested.31 If the Separation of Powers — and especially that involving the judiciary — is to be protected in its formal and symbolic importance, the coürts must be the guardians. But to do so we must — at least in part— overcome the functionalism that we were all taught.
The second reason for the errors that I believe the in banc court’s opinion makes stems from federal judges’ too ready assumption of their own capacity to ascertain state law. One of the great defects of diversity jurisdiction — which I tend to defend — is that all too often it gets federal judges into the business of determining what the law of the states is.32 And courts that in ordinary eases believe that they can easily decide what state contract law is may be tempted, as I believe the in banc court was today, to decide what is not for us to say — whether a state should or should not find that a contract exists.
The third reason most circuits have read the PLRA incorrectly has to do with what I think is an unfortunate view commonly taken of legislative history. I am not one who holds that words always or even frequently explain themselves. But that does not mean that specific pronouncements by legislators are free from even greater dangers as sources of legislative intent.33 All in all, when one must move beyond legislative language, it is often best to do so in the way that was traditional in the English courts. Barred from citing Hansard — the British parliamentary materials — the English courts focused on “the mischief’ that a statute sought to cure.34 Having identified the mischief, they read the words of the law in its clarifying light. Had the various circuits focused on the “mischief’ that the PLRA was trying to overcome, they would, I submit, have been far readier to accept a reading of the statute that barred the future effects — state or federal — of the decrees, but did not terminate the decrees themselves.
The final reason why, in my view, many judges get this case wrong stems from an error that is commonly made in dealing *192with statutory language. Thus, the in banc court readily acknowledges the ambiguities that characterize the operative language of the relevant part of the PLRA. But when the court looks at the definitional section of the law, it deems itself bound by the words used without considering the possibility — patently true in this case— that the definitions are themselves ambiguous. It is as if, when dealing with definitions, legislative language were necessarily clear and therefore had to be read as if it were, despite the manifest abuse of words that such a reading entails.
I would suggest, instead, that definitional parts of a statute, like the rest of the statute, must be given their ordinary meanings, when such meanings make sense, but must be read in their context, with a view toward the mischief the statute addresses, when the seemingly ordinary meanings lead to statements that are — to put it mildly — linguistically unnatural. The definition of “relief’ in the PLRA, which purports to include in relief “consent decrees” and also to exclude from relief “private settlement agreements,” makes no sense. It is bizarre because even if “consent decrees” might (awkwardly as the in banc court recognizes) be termed relief, “agreements” are simply not a form of relief. And hence to exclude them would be, if that is what the definition means, an exercise in surplusage and in futility.35 Read to make sense and read in the light of the mischief to be cured, the definitional section must mean that the relief to be barred in the future includes all relief available under or pursuant to consent decrees but does not include that relief which is available or pursuant to private settlement agreements. Such an elision would be readily assumed in reading the operative language of a statute if it were needed to make the statute make sense. Why should it not equally be read back into the definitional section of a law, when to do so accomplishes what .the statute was passed to do, preserves the formal boundaries between the branches of government, and makes the definition parse?
* * *
Because I believe the reading given to this statute by the in banc court distorts language, presumes a most unlikely congressional intent, improperly seeks to decide questions of state law, and, most important, does grievous damage to the formal lines that separate the coordinate branches of our government and does it to no purpose whatsoever, I respectfully dissent from most of the court’s reasoning. Because I agree with the court on why the case must be remanded to the district court, I join Part IV of its opinion. And because — despite what I believe to be a profoundly incorrect discussion — there is little in the court’s actual result that I would not also reach, I concur in its judgment.36
. Concededly, all the other courts of appeals that have confronted the legislation before us today have reached similar — or, in the case of a panel of the Ninth Circuit, which held the statute unconstitutional, equally unfortunate — results. See, e.g., Imprisoned Citizens Union v. Ridge, 169 F.3d 178 (3d Cir.1999); Taylor v. United States, 143 F.3d 1178 (9th Cir.1998), petition for rehearing granted, and panel opinion withdrawn, 158 F.3d 1059 (9th Cir.1998) (en baric); Hadix v. Johnson, 133 F.3d 940 (6th Cir.1998), cert. denied, - U.S. -, 118 S.Ct. 2368, 141 L.Ed.2d 737 (1998); Dougan v. Singletary, 129 F.3d 1424 (11th Cir.1997) (per curiam), cert. denied, - U.S. -, 118 S.Ct. 2375, 141 L.Ed.2d 743 (1998); Inmates of Suffolk County Jail v. Rouse, 129 F.3d 649 (1st Cir.1997), cert. denied, — U.S. -, 118 S.Ct. 2366, 141 L.Ed.2d 735 (1998); Gavin v. Branstad, 122 F.3d 1081 (8th Cir.1997), cert. denied, - U.S. -, 118 S.Ct. 2374, 141 L.Ed.2d 741 (1998); Plyler v. Moore, 100 F.3d 365, 372 (4th Cir.1996). On this point, see A Cautionary Note, infra Part III.
. Except in certain limited circumstances. See ante at 189-91.
. See infra note 21.
. See Benjamin v. Jacobson, 124 F.3d 162, 168, 174 (2d Cir.1997).
. The panel opinion, in contrast, argued that reading the PLRA’s definition of "prospective relief” found in § 3626(g)(7) (which the PLRA orders to be "immediately] terminat[ed]’’) to include consent decrees “has significant linguistic problems.” Benjamin II, 124 F.3d at 167. This was because decrees or settlement agreements are not forms of relief akin to injunctions or damages. Relief, instead, is generally understood to be made available under or pursuant to a consent decree or settlement agreement. Thus, although one could read the PLRA’s definition of "prospective relief" (which includes "all relief other than compensatory monetary damages,” 18 U.S.C. § 3626(g)(7), and “means all relief in any form that may be granted or approved by the court, and includes consent decrees but does not include private settlement agreements,” id. § 3626(g)(9)), to encompass consent decrees, the panel instead read the termination of prospective relief ordered under § 3626(b)(2) as leaving the decrees intact, but removing federal jurisdiction to order any future relief under those decrees. See Benjamin II, 124 F.3d at 167.
The panel did so because adopting the first reading would require viewing private settlement agreements as a type of relief. And that, the panel said, made no sense because private settlement agreements are simply not a form of relief. Therefore, the panel argued, one should "read the definitional phrase as saying 'the term "relief” means all relief in any form that may be granted or approved by the court, and includes [all relief granted pursuant to] consent decrees but does not include [relief granted pursuant to] private settlement agreements.’ ” /¿.(alteration in original). I continue to believe that this reading of the provision is the correct one.
To put it another way, while Congress can, for example, direct us to treat bananas as if they were apples, when Congress, in a definitional section, seems to say that bananas are apples, we should ask whether that is really what Congress meant or whether, instead, it merely intended to bring about the same result (identical treatment for both types of fruit) without requiring such an abuse of words.
. And here, the fact that the in banc majority reads the PLRA to bar prospective relief in state as well as federal courts becomes crucial. As the panel read the PLRA, the decrees — if they survived — could be enforced in state courts. Hence, under the panel's reading, the termination vel non of the decrees themselves directly affected results. It is for that reason that the panel said that the difference between termination of the decrees and termination of future relief under them was not simply formal. See Benjamin, 124 F.3d at 168. As the in banc court reads the PLRA, instead, future relief is barred in any court. That being so, the result is identical whether the decrees are terminated or whether courts are precluded from giving future relief under them.
. See Plaut v. Spendthrift Farm, Inc., 514 U.S. 211, 240, 115 S.Ct. 1447, 131 L.Ed.2d 328 (1995) (quoting Robert Frost, Mending Wall, in The New Oxford Book of American Verse 395-96 (Richard Ellmann ed., 1976)). Of course, Justice Breyer, who wrote separately in Plaut, was correct to point out that the poet was making a different point. See Plaut, 514 U.S. at 245, 115 S.Ct. 1447 (Breyer, /., concurring in the judgment) (quoting Frost in turn: “Before I built a wall I’d ask to know/ What I was walling in or walling out.”). But regardless of whether Justice Breyer is correct as a literary critic, Justice Scalia’s point is well-taken — the Framers had a keen appreciation of the importance of drawing sharp distinctions (building high walls, as it were) between the domains of the respective branches of our federal government.
. See Guido Calabresi, Retroactivity: Paramount Powers and Contractual Changes, 71 Yale L.J. 1191, 1202-03 (1962) (collecting cases, and noting that, even when a government is a party to a contract and alters that contract to its financial benefit, the government is not barred from doing so, as long as it demonstrates that the change was also passed in part for the general welfare).
. Subsequent Supreme Court cases have repeatedly interpreted Wheeling Bridge II as being premised on this rule. See, e.g., United States v. Chandler-Dunbar Water Power Co., 229 U.S. 53, 70, 33 S.Ct. 667, 57 L.Ed. 1063 (1913) (citing Wheeling Bridge II for the proposition that "[i]t is for Congress to decide what is and what is not an obstruction to navigation”); Gibson v. United States, 166 U.S. 269, 272, 17 S.Ct. 578, 41 L.Ed. 996 (1897) (noting that in Wheeling Bridge II, "it was ruled that the power of congress to regulate commerce includes the regulation of intercourse and navigation, and consequently the power to determine what shall or shall not be deemed, in the judgment of law, an obstruction of navigation”); see also United States v. Klein, 80 U.S. (13 Wall.) 128, 146-47, 20 L.Ed. 519 (1871) ("[In Wheeling Bridge //] the court was left to apply its ordinary rules to the new circumstances created by the act. In the case before us no new circumstances have been created by legislation." (emphasis added)); cf. Robertson v. Seattle Audubon Soc'y, 503 U.S. 429, 438, 112 S.Ct. 1407, 118 L.Ed.2d 73 (1992) (concluding that a revised environmental statute, which mentioned specifically two pending federal cases, did not violate the Separation of Powers because it "compelled changes in law, not findings or results under old law”).
. The majority emphasizes that it is not deciding the application of United States v. Klein, 80 U.S. (13 Wall.) 128, 20 L.Ed. 519 (1871), to this case, noting that plaintiffs chose not to pursue the Klein argument, which they made to the panel, before the in banc court. Accordingly, the question remains open. I believe that interpreting the PLRA provisions before us to mean what the majority reads them to mean renders the provisions violative of both Plaut and Klein. In Klein, the Supreme Court held that the Separation of Powers doctrine is violated when Congress prescribes a rule of decision for courts to follow without permitting courts to exercise their judicial powers independently. See id., 80 U.S. (13 Wall.) at 146-47. The PLRA, on the majority’s reading, not only reopens final judgments by directing federal courts to cancel existing decrees, it also appears to dictate the rule of decision for a distinct group of Article III cases in violation of Klein.
What is deeply troubling, however, is that, although the majority today purports to leave the Klein question open, by interpreting the PLRA in the manner that it has, it precludes a future panel of this Court (faced with another challenge to the PLRA in which a Klein argument is pursued) from interpreting the PLRA in a manner that would avoid the grave constitutional difficulties under Klein that inhere in the majority’s interpretation. Such a future panel will therefore be forced to address the Klein issues head on — and might perhaps be led to strike down the PLRA provisions that would clearly survive on the Benjamin panel’s reading of the law. The statute would then fall — irony of ironies — as a result of the majority's desire to apply a non-existent congressional intent in the case before us.
. Professors Tushnet and Yackle argue similarly:
Plaut's analysis suggests that Wheeling Bridge cannot stand for the broad proposition that Congress may alter any injunction whatsoever. That broad proposition would mean that the fundamental separation-of-powers principle articulated in Plaut does not apply to injunctive actions at all. But nothing in the Court's statement of the principle suggests why congressional interference with actions at law is more problematic than congressional interference with equitable remedies.
The Court in Plaut indicated the way out ... Federal Rule of Civil Procedure 60(b), which authorizes the federal courts to relieve parties from a final judgment for specified reasons ... [and] reflects and confirms the courts’ own inherent and discretionary power.
The Court’s emphasis on judicial discretion, which might be understood in this context as an essential attribute of the judicial power protected by Plaut’s separation-of-powers holding, explains the limits of Wheeling Bridge’s holding. In Wheeling Bridge the interstate commerce clause gave Congress the power to declare that the bridge did not interfere with interstate commerce.
Mark Tushnet & Larry Yackle, Symbolic Statutes and Real Laws: The Pathologies of the Antiterrorism and Effective Death Penalty Act and the Prison Litigation Reform Act, 47 Duke L.J. 1, 61-62 (1997) (footnotes and internal quotation marks omitted).
As Alexander Hamilton wrote in the Federalist Papers, “[the] legislature without exceeding its province cannot reverse a determination once made, in a particular case; though it may prescribe a new rule for future cases.” The Federalist No. 81, at 484 (Alexander Hamilton) (Clinton Rossiter, ed., 1961) (emphasis added).
. Though the majority’s approach and mine have the same effect on prisoners’ rights in the case before us, the difference between us may not be totally formal. The majority’s position may permit the legislature to order the precise response that the courts must make as a result of an alteration in the underlying law. Thus, the majority seems to countenance a legislative order that a decree- be vacated, when in fact termination of the judgment would equally well comport with the changed law. In the instant case, the issue is of no practical significance, but it might well be in other situations. It is possible that the majority’s holding means to be more limited still and that its decision allows Congress to order a specific change in a judgment only when that change is the sole way of adapting the judgment to the altered underlying law. If that is so, the absurdity of ascribing such an intent to Congress becomes even clearer, for then under System Federation, the courts are bound to make the identical change on their own.
. This, of course, is not to suggest that the Separation of Powers doctrine is only about formalisms. Separation of Powers notions may well make some results unconstitutional no matter how they are obtained. Thus, Congress presumably could not require Article III judges regularly to take on prosecutorial functions, regardless of whether it did so directly or indirectly.
. Nowhere is this exemplified more dramatically than in the Court’s Eleventh Amendment jurisprudence. See, e.g., Osborne v. Bank of United States, 22 U.S. (9 Wheat.) 738, 857, 6 L.Ed. 204 (1824) (positing the critical distinction between suing a state eo nomine or its officers for Eleventh Amendment purposes: "it is the party named in the record”). A person suing a state in federal court, therefore, must sue its officers as opposed to the state itself — even though the result is the same. Similarly, one who wishes to sue a state officer for money damages in federal court must label the suit as against the officer in his or her individual capacity; otherwise the claim is barred by the Eleventh Amendment. See, e.g., Edelman v. Jordan, 415 U.S. 651, 678, 94 S.Ct. 1347, 39 L.Ed.2d 662 (1974) (holding that the Eleventh Amendment bars claims for retroactive damages that run against the state treasury). Thus, the Eleventh Amendment protects the formal Sovereignty of the state by rules that in most cases do not alter results.
.The same approach has been consistently applied by the Supreme Court in other areas to preserve the delicate balance among the branches of government as well as between our federal and state governments. See, e.g., City of Boerne v. Flores, 521 U.S. 507, -, *181117 S.Ct. 2157, 2164, 138 L.Ed.2d 624 (1997) (striking down the Religious Freedom and Restoration Act, noting that “[Congress] has been given the power 'to enforce,’ not the power to determine what constitutes a constitutional violation”); Printz, 521 U.S. at -, 117 S.Ct. at 2378 (striking down portions of the Brady Act, which imposed certain regulatory obligations on state law enforcement officials, based in part on the fact that the law would effect a diminishment of the executive branch’s power to administer the laws); New York, 505 U.S. at 187-88, 112 S.Ct. 2408 (holding a federal hazardous waste law imposing obligations on states unconstitutional for compelling states “to enact or administer a federal regulatory program”); Gregory v. Ashcroft, 501 U.S. 452, 470, 111 S.Ct. 2395, 115 L.Ed.2d 410 (1991) (requiring a clear statement by Congress demonstrating its intent that the Age Discrimination in Employment Act apply to state judges); Bowsher v. Synar, 478 U.S. 714, 726, 106 S.Ct. 3181, 92 L.Ed.2d 583 (1986) (“Congress cannot reserve for itself the power of removal of an officer charged with the execution of the laws except by impeachment.”); INS v. Chadha, 462 U.S. 919, 958-59, 103 S.Ct. 2764, 77 L.Ed.2d 317 (1983) (striking down a single-house legislative veto of executive decisions respecting the suspension of alien deportation orders); id. at 964, 103 S.Ct. 2764 (Powell, J., concurring in the judgment) (finding that the exercise of the legislative veto with respect to six individuals was “clearly adjudicatory”). In Printz, the Court posited that the "separation of the two spheres is one of the Constitution's structural protections of liberty.” 521 U.S. at -, 117 S.Ct. at 2378. And it declared that the separation and independence of the branches provide a commensurate and equally important structural protection. See id. (citing Gregory v. Ashcroft, 501 U.S. 452, 458, 111 S.Ct. 2395, 115 L.Ed.2d 410 (1991); The Federalist No. 51, at 323 (James Madison) (Clinton Rossiter ed., 1961); The Federalist No. 28, at 180-81 (Alexander Hamilton)).
. See, e.g., Edward J. DeBartolo Corp. v. Florida Gulf Coast Bldg. & Const. Trades Council, 485 U.S. 568, 575, 108 S.Ct. 1392, 99 L.Ed.2d 645 (1988); Crowell v. Benson, 285 U.S. 22, 62, 52 S.Ct. 285, 76 L.Ed. 598 (1932); cf. Almendarez-Torres v. United States, 523 U.S. 224, -, 118 S.Ct. 1219, 1228, 140 L.Ed.2d 350 (1998) (”[T]hose who invoke the doctrine must believe that the alternative is a serious likelihood that the statute will be held unconstitutional. Only then will the doctrine serve its basic democratic function of maintaining a set of statutes that reflect, rather than distort, the policy choices that elected representatives have made. For similar reasons, the statute must be genuinely susceptible to two constructions after, and not before, its complexities are unraveled....”).
. This, of course, is only true in either instance where the need-narrowness-intrusiveness standard is not met. See 18 U.S.C. § 3626(b)(3).
. It is not, moreover, harder to say that such rights are abrogated if the decrees themselves are left untouched and only their future effects are affected than it is to say that they do not survive if the decrees are deemed terminated. See infra Part II.B.
. See, e.g., Lockerty v. Phillips, 319 U.S. 182, 187, 63 S.Ct. 1019, 87 L.Ed. 1339 (1943) (“[N]othing in the Constitution ... requires Congress to confer equity jurisdiction on any particular inferior federal court.”); Lauf v. E.G. Shinner & Co., 303 U.S. 323, 330, 58 S.Ct. 578, 82 L.Ed. 872 (1938) (upholding the provision of the Norris-LaGuardia Act limiting the ability of the federal courts to grant injunctive relief, and noting that ”[t]here can be no question of the power of Congress ... to define and limit the jurisdiction of the inferior courts of the United States”). But see Henry M. Hart, Jr., The Power of Congress to Limit the Jurisdiction of Federal Courts: An Exercise in Dialectic, 66 Harv. L.Rev. 1362, 1397 (1953) (positing constitutional limits on Congress’ control of federal court jurisdiction).
. While the panel disagreed, I do not think that the majority is clearly wrong is its holding. The words of the statute barring future relief under the decrees are not expressly limited to federal courts. And there is some legislative history indicating that Congress foresaw the possibility of attempts to obtain future relief under the decrees in state courts and wished to forbid it. See supra Part I.A. Together these are enough to make plausible the conclusion that future relief based on the decrees was meant to be made unavailable in either set of courts.
.Since the majority interprets the PLRA as directly requiring the termination of the decrees at issue while I read the statute as doing no more than changing the underlying law so as to bar prospective relief under the decrees, it may seem strange that I concur in the court's holding ”terminat[ing][the] forward-looking provisions of [the] consent decrees.” Supra at 168. But, in fact, even under my *184reading of the PLRA, the court's result remains appropriate.
In System Federation, the Supreme Court instructed the lower courts to respond to legislative changes in underlying laws by opening and amending judgments that were still in their control. It follows that after the PLRA has, on my interpretation, prohibited future enforcement of the consent decrees before us, the courts are bound to react to that change. They may in principle do so in one of at least three ways. First, they may decide to do nothing and, leaving the decrees in place, decline to enforce their forward-looking provisions. Second, they may do what the district court and various other circuits have done and conclude that the best judicial reaction to the prohibition on future enforcement is to vacate the decrees themselves. (The fact that the statute does not mention vacatur no more precludes that reaction than did the fact that the statute at play in Wheeling Bridge II never mentioned the previous injunction keep the Wheeling Bridge Court from dissolving that injunction.) Third, they may decide that the most appropriate response to the new law is for the courts themselves to order the forward-looking provisions of the decrees terminated.
The effect on prisoners’ rights of all three of these reactions is essentially the same, but the third solution strikes me as the most appropriate response. The statute expressly authorizes the parties to go to court and seek termination of such forward-looking provisions. See 18 U.S.C. § 3626(b)(2). And there seems to me to be no reason why the courts should not grant precisely this relief. Accordingly, my reading of the PLRA justifies a court order ''terminat[ing][the] forward-looking provisions of [the] consent decrees.” The majority believes that the statute itself commanded termination. The effect being the same, I concur in the court's result.
. See Taylor v. United States, 143 F.3d 1178, 1184 (9th Cir.1998). The Ninth Circuit has withdrawn the panel opinion in Taylor, and has reheard the case en banc. See Taylor v. United States, 158 F.3d 1059 (9th Cir.1998) (en banc) (withdrawing panel opinion and granting rehearing en banc).
.The majority's opinion, in this respect, is entirely advisory and is not needed for its holding. Nor can its foray be justified under the guise of correcting analogous dicta in the panel opinion. In the first place, the panel’s discussion of the possible survival of state law claims was not dicta. It was made necessary by the panel’s conclusion that, without the possibility that such state claims might survive, the PLRA might well be unconstitutional on various grounds. See, e.g., Benjamin, 124 F.3d at 176. The majority rejects these constitutional claims without hesitation. That, it has a perfect right to do. But once it has done so, any further discussion of possible state law actions becomes totally unnecessary. See Almendarez-Torres v. United States, 523 U.S. 224, -, 118 S.Ct. 1219, 1228, 140 L.Ed.2d 350 (1998). The most that might be justified, to avoid any lingering confusion arising from the panel’s analysis, would be a statement to the effect that "since the in banc court finds the statute to be constitutional regardless of the existence of possible state law claims, it expresses no opinion on the matter and should not be read as agreeing with the panel's analysis of such claims.” See infra notes 26, 28.
. Cf. Gregory v. Ashcroft, 501 U.S. 452, 464, 111 S.Ct. 2395, 115 L.Ed.2d 410 (1991) (requiring a clear statement from Congress before a federal statute will be read to apply to state officials).
. The panel opinion noted, for example, that the decrees in the instant case
have generated a judicially administered structure comprising over ninety related court orders and extending to more than thirty discrete areas of prison administration. The areas include the handling of detainees' mail and property, cell and body searches, maintenance of the physical plant, food service, and health and sanitary issues.... In 1982, pursuant to the agreement of the parties, a court monitoring agency called the Office of Compliance Consultants ("OCC”) was created. The OCC has monitored compliance with the Consent Decrees since that time.
Benjamin, 124 F.3d at 165.
. The Third Circuit has recently reached the same conclusion in Imprisoned Citizens Union v. Ridge, 169 F.3d 178 (3d Cir.1999):
If the Inmates have valid contractual claims that survive termination, such claims are "based solely upon ... [Pennsylvania] law”.... The Inmates are therefore free to pursue relief in the Pennsylvania courts. It is not our province to speak to the validity of any "claims arising under [Pennsylvania] law,” or to award relief therefor.
Id. at 190.
. The panel opinion, in noting that some of these decrees might constitute state law contracts that could be enforced in state courts, observed that a federal question would arise if state courts discriminated against those contracts embodied in federal consent decrees, or those that were entered into in exchange for the giving up of federal claims. See Benjamin, 124 F.3d at 178-79 & n. 23. That is surely correct, cf. Howlett v. Rose, 496 U.S. 356, 367-72, 110 S.Ct. 2430, 110 L.Ed.2d 332 (1990) (holding that state courts' refusal to entertain a § 1983 claim against a school district when state courts entertained similar state-law actions against state defendants violated the Supremacy Clause), but can be misunderstood. The point that the panel sought to make was not that state courts were bound to find valid state contracts embodied in the consent decrees. Far from it. All the panel said was that a state that held certain agreements to be valid state contracts could not discriminate against identical meetings of the minds and refuse to enforce them, simply because they were entered into in exchange for federal claims. State law can and must determine what it deems to be valid contracts and can do so fully and without the kind of interference that I suggest is implicit in the in banc court’s opinion. What state law cannot do, without raising significant federal questions, is to refuse to enforce one of two or more otherwise valid contracts simply because that contract involves federal rights.
. The Third Circuit, in dicta, twice asserted that the PLRA does not bar such potential state contract actions. See Imprisoned Citizens Union, 169 F.3d at 190 ("If the Inmates have valid contractual claims that survive termination, such claims are "based solely upon ... [Pennsylvania] law," and are not affected by the PLRA.... The Inmates are therefore free to pursue relief in the Pennsylvania courts."); see also id. at 187 n. 6 (“We express no opinion as to whether the Inmates have private rights in the consent decree. See infra, Section II.C.2. We simply note that if they do, those rights exist under state law and are not affected by the PLRA. See 18 U.S.C. § 3626(d) ('The limitations on remedies in this section shall not apply to relief entered by a State court based solely upon claims arising under State law.’).”). That Circuit appears to be as certain of this conclusion as the majority here, in dicta, is of the opposite result. The existence of these conflicting obiters underscores the wisdom of not trying to decide the issue until it is firmly presented in a case in which full arguments on both sides are made. See ante, concurrence of Judge LEVAL.
. See, e.g., Shaw v. Reno, 509 U.S. 630, 645, 113 S.Ct. 2816, 125 L.Ed.2d 511 (1993) (noting that, in voting rights cases, "district lines obviously drawn for the purpose of separating voters by race require careful scrutiny under the Equal Protection Clause regardless of the motivations underlying their adoption"); Griggs v. Duke Power Co., 401 U.S. 424, 429-30, 91 S.Ct. 849, 28 L.Ed.2d 158 (1971) (holding that, under Title VII, "practices, procedures, or tests neutral on their face, and even neutral in terms of intent, cannot be maintained if they operate to 'freeze' the status quo of prior discriminatory employment practices”).
. See also United States v. Then, 56 F.3d 464, 468 (2d Cir.1995) (Calabresi, concurring) ("If Congress, for example, though it was made aware of both the dramatically disparate impact among minority groups of enhanced crack penalties and of the limited evidence supporting such enhanced penalties, were nevertheless to act affirmatively and negate the Commission's proposed amendments to the Sentencing Guidelines (or perhaps were even just to allow the 100-to-I ratio [of sentences for crack cocaine to sentences for powder cocaine] to persist in mandatory minimum sentences), subsequent equal protection challenges based on claims of discriminatory purpose might well lie.").
. This was manifest during the in banc oral argument in the instant case, at which the plaintiffs expressed little concern with whether the decrees themselves, or only relief under them, would be terminated. The fact that for this reason parties rarely raise or argue such distinctions of course makes it harder for the courts to recognize them.
. To avoid that bad habit, I have argued that we should be much readier to certify questions of state law to state courts. See McCarthy v. Olin Corp., 119 F.3d 148, 157-61 (2d Cir.1997) (Calabresi, dissenting). But that practice remains relatively infrequent. See John B. Corr & Ira P. Robbins, Interjurisdic-tional Certification and Choice of Law, 41 Vand. L.Rev. 411, 418-31 (1988); accord Bradford R. Clark, Ascertaining the Laws of the Several States: Positivism and Judicial Federalism After Erie, 145 U. Pa. L.Rev. 1459, 1549 (1997) (observing that certification is discretionary and occurs on an ad hoc basis within the circuits).
. See, e.g., Green v. Bock Laundry Machine Co., 490 U.S. 504, 527-30, 109 S.Ct. 1981, 104 L.Ed.2d 557 (1989) (Scalia, J., concurring in the judgment).
. See, e.g., Heydon’s Case, 3 Co. Rep. 7a, 7b, 76 Eng. Rep. 637, 638 (1584) (noting that “the office of all the Judges is always to make such construction as shall suppress the mischief, and advance the remedy”); accord Gorris v. Scott, L.R. 9 Ex. 125 (1874); see also Millar v. Taylor, 98 Eng. Rep. 201 (1769) (first announcing the rule that courts may not consult Parliamentary material as an aid to statutory construction). Recently, amidst a heated debate, the English courts have (perhaps influenced by the American courts) begun to consult Hansard. See, e.g., Pepper v. Hart, 1 All E.R. 42 (1993).
. As I suggested earlier, a legislature can tell us to treat a banana the same way as an apple. But if it seems to say that a banana is an apple, we should not hesitate to try to figure out what it really meant to say.
. See supra note 21.